In a previous article we provided an overview of the different types of analytics one can run for problems related to customer acquisition, customer retention and customer churn. We mentioned how most of these questions either fall into strategic or tactical categories and can be addressed by either descriptive analytics or predictive analytics. In this article we will explore in a little more detail some of the tactical problems that can be addressed by customer lifetime value related predictive analytics.
1. Is it possible to tell from the data who would be ready to convert to becoming a customer and at what time?
Predictive analytics has had the longest history in marketing mainly to answer exactly these types of questions and all “Analytics Competitor” type companies today have mastered the use of analytics to answer this tactical question. However many small businesses in the marketing space today still simply make a random selection from their prospect list and call everyone in that selection!
The models which can help address these questions are called “Activation” or “Response” models. They will allow you to predict if a prospect will become a full-fledged customer. These models have been perfected in the financial services industry such as credit card or insurance. For example, for a credit card prospect to become an active customer, the prospect must respond to an offer, be approved for credit, and finally use the card. If the customer never uses the card, he or she actually may end up costing the bank more than a nonresponder! An insurance prospect can behave in much the same way. Just like in the case of credit card responses, a prospect can respond and be approved, but if he or she does not pay the initial premium, the policy is never activated and money is spent in getting them to the table without converting them.
Response and Activation models are actually two different models: one predicts response and another that predicts activation, given a response. The final probability of activation from the initial offer is the product of the probabilities predicted by these two models.
2. Can we predict who is ready to make a higher purchase? What has the most effect on getting a quick repeat purchase?
The models which answer these types of questions are called cross-sell and up-sell models. Cross-sell models are used to predict the probability of a current customer buying a different product or service from the same business. Up-sell models are used to predict the probability or value of a customer buying more of the same products (or services) or a higher valued product (or service). It is universally acknowledged that selling to current customers is far more cost-effective than new customer acquisition.
Companies are encouraged to do what is called “Testing offer sequences” which can help determine what is the best offer to make and when to make the next offer. This allows companies to carefully manage offers to avoid over-soliciting and possibly losing their customers. This brings us to the next closely related question.
3. Can we see what had an impact on donors who downgraded? Was it the amount of mailings, the time of the mailing, the type of mailing?
These questions call for the classic customer attrition or customer churn type models that predict the act of reducing or ending the use of a product or service after an account has been activated. Attrition is defined as a decrease in the use of a product or service. For example in credit card industry, Attrition may be defined as the decrease in balances on which interest is earned. Churn is defined as the closing of one account in combination with the opening of another account for the same product or service, usually at a reduced cost to the consumer and usually with a competitor company. This is a major problem in another industry: the telecom business.
4. What can we say about the average life time that a customer stays with the business?
This brings us to the last question which is in reality more of a strategic rather than tactical question. A customer lifetime value related model attempts to predict the overall profitability of a customer (person or business) over a predetermined length of time. We have several detailed articles on how to develop customer lifetime value (CLV) models and will not repeat here. One thing we do need to mention is the data needed for any of the above models discussed. All of them requires one or more types of data: demographic, behavioral and attitudinal. The nature of these and some examples will be discussed in another article.
Originally posted on Fri, Nov 15, 2013 @ 06:40 AM